1. AI Agents Are Terrible Freelance Workers
A sobering study finds that even the most advanced AI agents manage to complete less than 3 % of simulated freelance tasks—video editing, graphic design, data scraping—earning just $1,810 out of a possible $143,991, puncturing the myth that AI will soon replace human freelancers.
“Workers may use AI as a tool to boost productivity, but agents on their own remain ill-suited for complex, multi-step work without long-term memory or on-the-job learning.” (wired.com)
2. Human Centric design is empathy and taste over LLM output
In an age of tool-driven workflows, true UX mastery still lies in the designer’s craft, empathy, and discerning taste—not merely in prompting generative models.
“Craft + Empathy + Taste = Human-Centric Design” (medium.com)
3. Can AI tell us anything meaningful about Bob Dylan’s songs?
A thoughtful exploration of machine analysis on Dylan’s corpus reveals that while algorithms can map structural shifts, they cannot penetrate the emotional mysteries that make his work resonate.
“Despite the revelations, Dylan’s cultural power lies in remaining, even now, a complete unknown.” (aeon.co)
4. 8bit-GPT: Exploring Human-AI Interaction on Obsolete Macintosh Operating Systems
Reviving the charm of 1980s Mac interfaces, 8bit-GPT deliberately throttles AI assistance to defamiliarize the user, spotlighting how friction and “slow tech” can deepen our awareness of machine agency.
“By foregrounding inefficiency and counterfunctionality, the project challenges prevailing narratives of seamless AI integration.” (arxiv.org)
5. The Photographer Eye: Teaching Multimodal Large Language Models to See and Critique like Photographers
Introducing PhotoCritique, a new dataset of expert critiques, and PhotoEye, a model that fuses multi-view vision with linguistic guidance to emulate photographic aesthetic judgment—bridging AI research with visual artistry.
“By capturing photographic techniques, composition rules, and post-processing insights, PhotoEye advances multimodal aesthetic understanding.” (arxiv.org)